This year's problem seems simple. You are given a map of a city and have to code a "robber bot" which steals money from the banks and a "cop bot" which coordinates with other cops to catch the robber. But of course, it gets more complicated than that - read the spec in detail and you find that the big problem is full of little problems like speaking the protocol, path finding, maximising scores, making plans for the other cops , trust models for cop bots, voting on plans, and so on and so on. It's a good problem.
We had a small team - just me and Greg McCarroll, mostly due to Glastonbury and YAPC::NA clashes. We parsed and talked the protocol fairly quickly, had static and random robbers and cops shortly after that and then started with the algorithms. The cycle went something along like this: modify the cop so that it catches the robber, then modify the robber so that it escapes the cops, then repeat. We managed a submission a few hours before the deadline, and one a few minutes before it too
I regret the small size of the team. It's really handy having extra people there to bounce ideas off, and so that they can take care about an aspect of the problem. For example, we only added visualisation late on Sunday and that helped a lot for tweaking parameters. I have some animated GIFs to show off:
Is "Team Orange" going to win? Probably not, but it was fun to enter!